2000 Fiscal Year Final Research Report Summary
Studies on Two-Dimensional Warping
| Project/Area Number |
10680385
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Kyushu University |
|---|
Principal Investigator |
SAKOE Hiroaki Graduate School of Information Science and Electrical Engineering, Kyushu University Professor, 大学院・システム情報科学研究院, 教授 (30215701)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
UCHIDA Seiichi Graduate School of Information Science and Electrical Engineering, Kyushu University Research Associate, 大学院・システム情報科学研究院, 助手 (70315125)
KATAYAMA Yoshinori Graduate School of Information Science and Electrical Engineering, Kyushu University Research Associate, 大学院・システム情報科学研究院, 助手 (00214338)
KAMATA Sei-ichiro Graduate School of Information Science and Electrical Engineering, Kyushu University Associate Professor, 大学院・システム情報科学研究院, 助教授 (00204602)
|
|---|
| Project Period (FY) |
1998 – 2000
|
| Keywords | Two-Dimensional Warp / Tological Consistency / Continuity / Monotonicity / Dynamic Programming / Beam Search / Elastic Matching / Piecewise Linear approximation |
|---|
| Research Abstract |
Two-dimensional warping between images was investigated, in which an optimal pixel to pixel mapping is searched for by Dynamic Programing (DP).
(1) Basic study was made on topological consistency of two-dimensional warp, with conclusion that two-dimensional monotonic and continuous constraints give a good approximation of topological consistency of the warp. An minimization problem was formulated which defines an optimality of warp, with residual error measure as the objective function and monotonicity and continuity as constraints.
(2) Optimization algorithm was investigated for the above defined minimization problem. DP was succsseffully applied, giving an O(9^N) time-complexity algorithm for N×N image warping. Afurther acceleration was achieved by use of beam search technique.
(3) An approximation algorithm was proposed in which piecewise linear warp is attained by optimizing limited number of control points. By this approximation a polynomial order algorithm was established, with a slight degradation in warp accuracy.
(4) Proposed algorithms were experimentally applied to handwritten character recognition problems. Their potential applicabilities were demonstrated. Problems necessitating further investigation were clarified.
|
Research Products
(41 results)
